If you want a book that's a bit more modern than A Random Walk Down Wall Street, try David Aaronson's book, Evidence-Based Technical Analysis: Applying the Scientific Method and Statistical Inference to Trading Signals. The author discusses a lot of the evidence for and against certain technical indicators; for example, he discusses how head-and-shoulders indicators are usually worse than random when making trading decisions. I don't have a copy of the book handy, so you'll have to read it and decide for yourself based on the evidence, but it's a more recent and rigorous empirical look at TA.
If you're interested in academic studies, a good starting point is Foundations of Technical Analysis: Computational Algorithms, Statistical Inference, and Empirical Implementation by Lo, Mamaysky, and Wang. Their study finds that several technical indicators, e.g. head-and-shoulder, double-bottom, and various rectangle techniques, do provide marginal value. They also find that although
human judgment is still superior to most computational algorithms in the area of visual pattern recognition, ... technical analysis can be improved by using automated algorithms
Since this paper was published in 2000, computing power and statistical analysis have gained significant ground against human abilities to recognize, and more importantly exploit, technical indicators. This statement isn't necessarily a criticism of TA, but rather a warning against its use by the average day trader in markets where computerized traders are highly active, since high-speed traders can identify and take advantage of signals faster than retail day traders, thus potentially decreasing or removing the potential to gain from the signals.
This applies to a lesser extent in markets where high-speed trading isn't as active or over longer time periods; for longer time periods, however, it's important to understand that some signals (like head-and-shoulders) were found to be detrimental to trading returns. Some anecdotal evidence suggests otherwise, but a) this is anecdotal evidence, and b) even if a signal has a completely random effect on trading returns, there will always be some traders who earn a positive return with it, either on its own or in conjunction with others; such is the nature of statistical distributions. It's why collecting data and backtesting strategies is more important than online testimonials and similar sources (including this post). There are always statistically fortunate people; see my story in this post about pamphlets that predict sports results.